Examples of org.apache.uima.ruta.textruler.core.TextRulerExampleDocument

Package org.apache.uima.ruta.textruler.core

Examples of org.apache.uima.ruta.textruler.core.TextRulerExampleDocument

org.apache.uima.ruta.textruler.core.TextRulerExampleDocument
TextRulerExampleDocument stands for one document usually loaded from an XMI file. It uses the given CasCache for storing its CAS with the XMI filename as the key. It holds ArrayLists for positive and negative MLExamples which can be filled on demand for a given learning target. E.g. single slot algorithms learn rules for each slot separately, so the work-flow is to clear the current examples and create new for the next slot target. The same is with single slot boundary algorithms like LP2: It first creates all left boundary examples, learns from them, clears the examples and creates the right boundary examples and so on. This class also provides the functionality extract and created MLExmaples of a given document or test CAS for a given TextRulerTarget. Especially for boundary algorithms you can call createBoundaryAnnotationsForCas to get boundary annotations at the beginnings and endings of an example slot. Caution (this is quite a bit inconvenient at the moment!): If a CAS gets loaded from the casCache, you have to call createBoundaryAnnotationsForCas again, so your casLoader must be aware of that (see BasicLP2 for an example) ! hint: this could be renamed to MLDocument instead of TextRulerExampleDocument ?

    interTupelSeparatorsCache.clear();
    TextRulerToolkit.log("-- WIEN END");
  }


  protected boolean findRightPatterns() {
    TextRulerExampleDocument doc = exampleDocuments.getDocuments().get(0);
    boolean allFound = true;
    for (int k = 0; k < slotNames.length; k++) {
      List<TextRulerRulePattern> rightContexts = getRightContextForSlot(doc, k);
      System.out.println(rightContexts.get(0));
      int shortest = Integer.MAX_VALUE;

View Full Code Here

    }
    return allFound;
  }


  protected boolean findLeftPatterns() {
    TextRulerExampleDocument doc = exampleDocuments.getDocuments().get(0);
    // skip l 0 !
    if (slotNames.length < 2)
      return true;
    boolean allFound = true;
    for (int k = 1; k < slotNames.length; k++) {

View Full Code Here

    return allFound;
  }


  protected boolean findHeadTailAndL1Patterns() {
    List<TextRulerExampleDocument> docs = exampleDocuments.getDocuments();
    TextRulerExampleDocument doc0 = docs.get(0);
    TextRulerRulePattern head = new TextRulerRulePattern();
    TextRulerRulePattern tail = new TextRulerRulePattern();
    getPageHeadAndTailPortion(doc0, head, tail);


    final class HLCandidate {

View Full Code Here

   * @return ranked list of conditions
   */
  private RankedList createConditions(TrabalAnnotation annotation) {
    RankedList result = new RankedList(idf);
    TrabalRuleItem ruleItem = new TrabalRuleItem(annotation);
    TextRulerExampleDocument doc = annotation.getDocument();
    CAS cas = doc.getCAS();
    for (TrabalRuleItem item : getTermsBefore(ruleItem, doc)) {
      result.add(new Condition(ConditionType.AFTER, item));
    }
    for (TrabalRuleItem item : getTermsAfter(ruleItem, doc)) {
      result.add(new Condition(ConditionType.BEFORE, item));

View Full Code Here

      if (inducedRules.containsKey(ruleString)) {
        rule.setCoveringStatistics(inducedRules.get(ruleString));
        System.out.println("skipped with " + inducedRules.get(ruleString));
      } else {
        for (int i = 0; i < goldDocs.size(); i++) {
          TextRulerExampleDocument goldDoc = goldDocs.get(i);
          TextRulerExampleDocument additionalDoc = additionalDocs.get(i);
          sendStatusUpdateToDelegate("Testing " + ruleSet + ruleInfo + " on document " + (i + 1)
                  + " of " + goldDocs.size() + " : rule " + counter + " of " + rules.size(),
                  TextRulerLearnerState.ML_RUNNING, false);
          TextRulerStatisticsCollector sumC = new TextRulerStatisticsCollector();
          prepareTestCas(theTestCAS, goldDoc, additionalDoc);

View Full Code Here

   * @param truth
   *          The truth example
   */
  private List<TrabalRule> createShiftAllRules(TextRulerExample error, TextRulerExample truth) {
    List<TrabalRule> result = new ArrayList<TrabalRule>();
    TextRulerExampleDocument document = error.getDocument();
    CAS cas = error.getDocumentCAS();
    List<TrabalRuleItem> truthLeftBorder = getBorderOfExample(truth, document, cas, true);
    List<TrabalRuleItem> truthRightBorder = getBorderOfExample(truth, document, cas, false);
    List<TrabalRuleItem> errorLeftBorder = getBorderOfExample(error, document, cas, true);
    TextRulerTarget target = error.getTarget();

View Full Code Here

   * @param truth
   * @return list of expansion rules
   */
  private List<TrabalRule> createExpansionRules(TextRulerExample error, TextRulerExample truth) {
    List<TrabalRule> result = new ArrayList<TrabalRule>();
    TextRulerExampleDocument document = error.getDocument();
    CAS cas = error.getDocumentCAS();
    boolean shiftToLeft = error.getAnnotation().getBegin() > truth.getAnnotation().getBegin();
    List<TrabalRuleItem> border;
    if (shiftToLeft) {
      border = getBorderOfExample(truth, document, cas, true);

View Full Code Here

   * @return list of annotation rules
   */
  private List<TrabalRule> createAnnotationRules(AnnotationError each) {
    List<TrabalRule> result = new ArrayList<TrabalRule>();
    TextRulerExample truth = each.getTruth();
    TextRulerExampleDocument document = additionalDocuments.getDocuments().get(
            exampleDocuments.getDocuments().indexOf(truth.getDocument()));
    CAS cas = document.getCAS();
    List<TrabalRuleItem> truthLeftBorder = getBorderOfExample(truth, document, cas, true);
    List<TrabalRuleItem> truthRightBorder = getBorderOfExample(truth, document, cas, false);
    TextRulerTarget target = truth.getTarget();
    TrabalAnnotation truthAnnotation = (TrabalAnnotation) truth.getAnnotation();
    for (TrabalRuleItem front : truthLeftBorder) {

View Full Code Here

   * @return ranked list of conditions
   */
  private RankedList createConditions(TrabalRuleItem frontBoundary, TrabalRuleItem rearBoundary,
          TrabalAnnotation truth) {
    RankedList result = new RankedList(idf);
    TextRulerExampleDocument doc = additionalDocuments.getDocuments().get(
            exampleDocuments.getDocuments().indexOf(truth.getDocument()));
    CAS cas = doc.getCAS();
    for (TrabalRuleItem item : getTermsBefore(frontBoundary, doc)) {
      result.add(new Condition(ConditionType.AFTER, item));
    }
    for (TrabalRuleItem item : getTermsAfter(rearBoundary, doc)) {
      result.add(new Condition(ConditionType.BEFORE, item));

View Full Code Here


    trainingDocuments = new ArrayList<TextRulerExampleDocument>();
    CasCache casCache = new CasCache(5, this);


    for (String fileName : getXMIfileNames(trainingFolder))
      trainingDocuments.add(new TextRulerExampleDocument(fileName, casCache));
    testFileNames = getXMIfileNames(testFolder);
  }

View Full Code Here

0 1

TOP

Related Classes of org.apache.uima.ruta.textruler.core.TextRulerExampleDocument

org.apache.uima.analysis_engine.AnalysisEngine

org.apache.uima.cas.CAS

org.apache.uima.cas.FeatureStructure

org.apache.uima.cas.text.AnnotationFS

org.apache.uima.cas.Type

org.apache.uima.cas.TypeSystem

org.apache.uima.ruta.textruler.learner.trabal.TrabalLearner

org.apache.uima.ruta.textruler.learner.wien.Wien

org.apache.uima.ruta.textruler.tools.F1Scorer

org.apache.uima.util.CasCopier

All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.